Henry Kissinger speaks at College of Computing celebration
Kissinger: ‘transparency that was essential for arms control’ hard to establish for AI
Henry Kissinger spoke in “How the Enlightenment Ends,” a discussion moderated by Thomas Friedman at the College of Computing celebration Feb. 28 in Kresge Auditorium. Kissinger published an eponymous piece about AI in The Atlantic in June 2018.
MIT’s invitation of Kissinger, who served as national security advisor and secretary of state during Richard Nixon and Gerald Ford’s presidential administrations, has been a subject of controversy among the MIT community. During Kissinger’s talk, a protest was held nearby on the steps of the student center. A banner above the protest read, “Celebrating Complicity with a Billionaire and War Criminal.”
In the talk, Kissinger argued that machines aren’t governed by ethical or philosophical norms. “Right now, technology is way ahead of the humanists,” said Kissinger during the talk. “You have science … without having a philosophical framework within which to put [technology].”
Kissinger’s Atlantic article cited an AI chatbot named Tay which “proved unable to define the imperatives of ‘friendly’ and ‘reasonable’ language installed by its instructors and instead became racist, sexist, and otherwise inflammatory in its responses.”
Kissinger said during his talk that he became interested in AI after going to a conference in 2016. At the conference, Eric Schmidt, best known as former executive chairman of Google and later Alphabet Inc., encouraged him to attend a talk on the subject. Kissinger said he attended a talk on AlphaGo, the first computer program to beat a professional human Go player.
Kissinger expressed bewilderment at how the AlphaGo machine could, “by playing with itself, come up with a style of chess [sic] that has never been played before.”
“The input was something everybody had seen, and the output was something that had never been seen before,” Kissinger said at the talk. “We need to understand that process.”
Kissinger discussed his reaction upon learning that some self-driving cars inch forward in traffic at a stoplight: “Where do they learn that? That’s not in their instructions.” (Such cars are either directly programmed to do so, or otherwise would have learned to do so from training data on other drivers doing the same thing.)
Friedman asked Kissinger if, based on his experience with nuclear weapons regulations, he saw a similar need to regulate AI. Kissinger replied that arms control was a “theory of numbers” based on reducing the number of arms each country had. However, he said “the transparency that was essential for arms control would be very hard to establish” in AI and sharing information will become more complicated; therefore, “the other side’s ignorance is one of your best weapons.”
Friedman also asked Kissinger if he thought AI was the “end of democracy,” citing countries such as Iran and China which are gathering massive amounts of data on their own populations. Kissinger responded, “AI makes it possible to supervise your population … and to do it instantaneously.” Kissinger worried “how to keep good ideas from being dominated by dangerous ideas.”